Unrestrictable is an AI chat in the browser that runs open-weight models with the refusal behaviour removed or tuned out. You can try it with no account, nothing to install, and no card. This post explains what the free tier actually is, what is kept, and where the limits are. Figures are as the site stood on 2026-11-20.
What "no sign-up" means here
Open unrestrictable.ai/chat and type. There is no email field and no login; a human check stands in for the sign-in.
- Three free turns a day with no account.
- 239 models take those free turns, including GLM-5.3 Uncensored, DeepSeek V4 Pro Uncensored, GLM-5.3 Flash Abliterated, GLM-5.3 Flash Uncensored and DeepSeek V4.1 Flash Abliterated.
- 385 of the site's 396 models are in the chat. The rest of them open with credit.
The model picker sits in the composer. Each model is marked Free to try or With credit, and the free ones come first. You can switch models mid-conversation: the next reply comes from the one you picked and the earlier turns stay.
What the models are
The catalog is open-weight builds in three kinds:
- Abliterated: the refusal direction is removed from the weights.
- Uncensored: tuned without the filtering a general assistant ships with.
- Fiction-tuned: trained for long-form writing and roleplay.
Every model gets the same test prompt daily and the result is public on the proof page; on 2026-11-20, 310 of 338 checked models passed.
What else the free chat does
- Files: attach a PDF, document or image up to 7 MB and it is read to text for the model.
- Web search: switch on Search the web and the reply cites the pages it found.
- Settings: a character, a system prompt and the temperature.
- Characters: 14 house characters, each a persona and an opening line. Every character is an adult.
- Voice input: the mic uses your browser's own speech recognition.
A file or a search counts as one of the day's free turns.
What is kept
This is the part worth reading before you type anything personal.
- Your conversation is saved in your browser and nowhere else.
- Each turn is one request through the gateway, which keeps a usage record for billing: the model, the token counts and the charge. It does not store the prompt or the reply.
- The company running the model receives the request so it can answer.
- New messages are classified for safety by a moderation API; when it raises a flag, the categories are kept, never the text.
- With voice input, your browser's speech recognition may send the audio to the browser's maker; the site receives only the text.
- Memory, which stores notes on the server side, starts off.
The privacy page lists every record.
"Unrestricted" does not mean "no rules"
The models answer without a stock refusal layer, so fiction can deal with dark subject matter and questions a filtered assistant declines get an answer. The acceptable use policy still applies to every request. It rules out, among other things, any sexual content involving minors, sexual content of a real person without their consent, attacking systems you are not authorized to test, and weapons of mass harm. On the image and video models, a prompt that describes a minor is refused before it reaches a model.
When the free turns run out
You will see a prompt for a key when today's free turns are used, the free budget is spent, or the model you picked runs on credit. Three options:
- A free account: 5 messages, 2 images and 3 files a day in the console chat, plus replies read aloud by a voice model, your own characters, a Create page for images, video, speech and music, and Memory if you switch it on.
- A top-up: a one-off payment from $10 by card, WeChat Pay, Alipay or crypto, with no ID and no phone number. 100 credits are $1, credits last 12 months, and each reply settles at that model's price per token on the pricing page.
- A usage plan that renews by card: Plus is $10 a month for $20 of usage, Pro $25 for $50, Max $100 for $200, Ultra $250 for $500. You can switch it off any time in Billing.
Using the same models from code
The chat is a front end on an API. The same models answer at an OpenAI- and Anthropic-compatible base URL, so SillyTavern, Cursor, Cline, OpenCode, Claude Code or the OpenAI SDK can use them with a key:
export OPENAI_BASE_URL="https://api.unrestrictable.ai/v1"
export OPENAI_API_KEY="your-key"
Paste a key from the console into the chat and every model there opens as well.
Try it
unrestrictable.ai/chat: three turns, no account. The models page lists all 396 with their prices.
Top comments (0)